Fall Guardian: An Intelligent Fall Detection and Monitoring System for Elderly

With the boom in the elderly population worldwide, the number of fall detection systems have greatly increased. With intelligent detection and monitoring, timely medical help can be ensured. Throughout the last decade, various systems were developed to detect falls accurately and notify caregivers timely. But most systems were developed based on sensor data of specific quality of sensors. As a result, those systems worked poorly on devices with different qualities of the same sensors. In this paper, we propose a system of fall detection and fall monitoring, Fall Guardian, which can be implemented in android devices with various sensor qualities. In this system, we sent collected sensor data to a cloud server with a Random Forest classifier and several thresholds for preprocessing, feature extraction, and prediction. Then a post-fall movement detection algorithm checked the severity of fall injury and alerted caregiver of elderly user with the elderly’s GPS location. Then we tested this application on multiple devices. The proposed system reached the highest accuracy of 98.32% and 96.64% with and without gyroscope sensor data. Environmental tests performed on twelve test subjects using four different smartphone models yielded an overall accuracy of 95.83%, a sensitivity of 93.33%, and a specificity of 98.33%.